In interviews about these systems, White has said that work is changing at three levels: individuals, workflows and business models. On an individual level, professionals can take on more complex work because AI takes over many of the rote tasks that previously filled their days. Product managers, for example, are already using agents to perform data‑science tasks that once required dedicated specialists.
At the workflow level, companies are tearing down long, sequential processes in areas such as marketing, compliance and product development. The near‑instant combination of internal data, external research and customer feedback can compress weeks of analysis into minutes of execution. That has clear implications for HR teams under pressure to deliver new policies, learning paths or workforce plans with fewer people and less time.
The third level – business models – may be the most significant for HR leaders thinking about the future of work, according to a report in the Wall Street Journal. If AI makes it far easier to test and launch new products or services, organisations can move into adjacent markets without building large new departments first. That can reduce the urgency of some hiring and shift HR’s focus toward developing adaptable, cross‑functional talent that can work effectively with agents.
Other leaders stress that AI’s value is less about replacing professionals and more about enabling them. Jeff Stibel, chief executive of LegalZoom, has said that AI makes it easier to start a business and that “AI provides the insight, but LegalZoom provides the trusted solution.” For HR, that distinction is important: AI can surface patterns in engagement data or suggest wording for policies, but employees will still look to human leaders for judgement, context and trust.
The architecture behind long‑running agents also alters which skills are most valuable on a team. Alex Salazar, co‑founder of AI startup Arcade.dev, has described how developers used to hand‑craft prompts such as “you are an accounting agent; here is the enterprise resource planning tool.” Now, users can assign higher‑level goals like “check depreciation schedules,” and the agent will “dynamically decide which skills and tools it needs to solve the problem, iterates on the plan, and executes it without rigid preprogramming.”